细化Transformer网络的弱监督图像语义分割  

Refining Transformer for weakly supervised image semantic segmentation

作　　者：孙万春 冯欣[1,2] 马慧 胡立松 Sun Wanchun;Feng Xin;Ma Hui;Hu Lisong(School of Computer Science&Technology,Changchun University of Science&Technology,Changchun 130022,China;Chongqing Research Institute,Changchun University of Science&Technology,Chongqing 401122,China;Computer Basic Teaching&Research Dept.,Anhui Vocational College of Police Officers,Hefei 230031,China;R&D Dept.,Beike Tianhui(Hefei)Laser Technology Co.,Ltd.,Hefei 230041,China)

机构地区：[1]长春理工大学计算机科学技术学院,长春130022 [2]长春理工大学重庆研究院,重庆401122 [3]安徽警官职业学院计算机基础教研室,合肥230031 [4]北科天绘(合肥)激光技术有限公司研发部,合肥230041

出　　处：《计算机应用研究》2023年第11期3515-3520,共6页Application Research of Computers

基　　金：安徽省自然科学研究重点资助项目(KJ2021A1471)。

摘　　要：图像级标签的弱监督图像语义分割方法是目前比较热门的研究方向,类激活图生成方式是最为常用的解决该类问题的主要工作方法。由于类激活图的稀疏性,导致判别区域的准确性降低。针对上述问题,提出了一种改进的Transformer网络弱监督图像学习方法。首先,引入空间注意力交换层来扩大类激活图的覆盖范围;其次,进一步设计了一个注意力自适应模块来指导模型增强弱区域的类响应;特别地,在类生成过程中,构建了一个自适应跨域来提高模型分类性能。该方法在Pascal VOC 2012验证集和测试集上分别达到了73.5%和73.0%。实验结果表明,细化Transformer网络学习方法有助于提高弱监督图像的语义分割性能。The weakly supervised methods for image semantic segmentation using the image-level labels are a relatively popular research direction.The class activation maps generation approach is the most commonly used approach in these researches.Due to the sparsity of class activation maps,the accuracy of discriminative regions is generally low.To address these problems,this paper proposed an improved weakly supervised image learning method based on the Transformer network.Firstly,it introduced a spatial attention interaction layer to extend the coverage of the class activation maps.Secondly,it designed an attention adaptive module to guide the model to enhance the class response in weak regions.In particular,it constructed an adpative cross domain to improve the model classification performance during class generation.The method achieves the accuracies of 73.5%and 73.0%on the Pascal VOC 2012 validation sets and test sets,respectively.The experimental results prove that the refined Transformer network learning method can improve the image semantic segmentation performance of weakly supervised method.

关 键 词：深度学习 弱监督学习 图像语义分割 TRANSFORMER 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]